Loading...

Abstract

Domain

ROBOTICS

Title

Machine learning based replicator robot

Abstract

The paper introduce a descriptive framework for robotic replicating systems that extends von Neumann's classical model of self-reproducing automata. A new physical prototype is developed to examine principles and uncover hardware limitations of kinematic reproduction at a low-complexity level in man-made physical systems. The initial functional robot assembles six subsystems located in a partially structured environment to form an exact functional replica of itself. Self-replication of the sort demonstrated here can be achieved by designing a robotic system composed of multiple simple parts/subsystems rather than a relatively few complex robotic modules. Therefore, this prototype hints at the feasibility of the concept of fully autonomous man-made machines that can construct functional copies of themselves from many very basic components. As a simple measure of system complexity, the number of active elements in each subsystem and interconnections between subsystems are counted. In addition, we present a measure of the degree of replication that includes the complexity distribution and the relative complexity of the total system to individual parts.